Review CVE reproduction and remediation tasks built to train a frontier lab's models, judging whether the exploit actually fires, the fix actually closes it, and the Docker environment actually recreates the vulnerable conditions. Output is rubric-scored written feedback on each task, at roughly full-time hours, remote and US-based.

Summary, eligibility guidance and screening preparation are prepared by Sidequest from the platform's posting. Check the source listing before applying.

What the work involves

You receive task bundles: a target CVE, a containerized environment meant to reproduce it, a proof-of-concept or exploit trigger, a proposed patch, and a two-part verification suite — functionality tests that confirm the application still works, and vulnerability tests that confirm the flaw is genuinely closed. Your job is to decide whether the whole package holds up. That means reading the upstream advisory and patch commit, standing up the Docker Compose stack, confirming the exploit path is the real one rather than an unrelated crash or a hardcoded shortcut, and checking that the fix addresses the root cause instead of filtering the single payload the test happens to use.

Most defects you catch will be subtle: a vulnerability test that passes against the patched build for the wrong reason, a container pinned to a dependency version that isn't actually affected, a remediation that sanitizes at the wrong layer, a CWE mapping that names the symptom rather than the weakness. You write this up against a rubric — dimensions like reproduction fidelity, remediation soundness, verification rigor, and environment accuracy — with concrete pointers to file, line, and test case, plus the specific change that would make the task acceptable.

What the screen looks for

Verifiable hands-on history. Named vulnerability classes you've exploited and fixed, not a list of tools. Expect follow-ups that go one layer deeper than your first answer.
Root-cause reasoning. Whether you can distinguish a fix from a patch-shaped workaround, and explain why a given sanitization point is or isn't sufficient.
Verification design judgment. Whether you understand why a vulnerability test must fail on the vulnerable build and pass on the patched one, and what it means when it doesn't.
Container fluency. Real experience composing multi-service reproduction environments, including networking, pinned versions, and non-determinism.
Reviewer temperament. Feedback that is specific and actionable rather than a verdict.

Logistics

Remote and US-based, contract, structured around roughly 40 hours per week. Review is largely asynchronous — you pull tasks from a queue and submit written assessments — though calibration sessions and rubric updates are common early on. Observed pay for this listing is $70–$90/hour; actual rate depends on assessed depth and is set by the platform, not guaranteed. Expect a paid or unpaid sample review as part of onboarding.

CVE Vulnerability Expert

What the work involves

What the screen looks for

Logistics

Interview prep — 10 questions